A Generalized Network Generation Approach for Agent-Based Models

For their initialization, many agent-based models require a population which corresponds in its essential characteristics to the examined real population. It should reflect the real distribution of attributes of interest, e.g., age, gender, or income, as well as the social network between the agents. Since a disaggregated data set with all required information is rarely available, a synthetic population must be created. Methods that assign realistic attribute values to agents are well studied in the literature. In contrast, the generation of plausible social networks has been less extensively researched, but several comprehensive adhoc models have been developed. The focus of this work is to introduce a reusable, generalized approach for the generation of synthetic social networks. Symbolic regression is used to automatically train generation rules based on a network sample, instead of having to define rules a priori. Manually specified constraints are taken into account to avoid implausible relationships.